GoKawiilOpinion: AI researchers argue LLMs lack true reasoning, unlike AlphaGo's move 37
An opinion piece argues that large language models operate like fast, associative 'System 1' thinking, predicting tokens without genuine step-by-step reasoning. The authors contrast this with AlphaGo's famous move 37 against Lee Sedol, which they say resulted from deliberate tree search rather than intuition alone, combining a policy network's hunches with explicit lookahead through thousands of possible game branches.
GoKawiil's interpretation of the reporting above, not reported fact.
The distinction matters because trustworthy AI outputs in fields like science and medicine may require deliberative reasoning, not just pattern completion, according to the authors. This suggests current LLM architectures could fall short of producing truly novel insights unless they incorporate search-like reasoning mechanisms similar to AlphaGo's.
- AlphaGo combined a 'System 1' policy network with a 'System 2' search mechanism to find move 37.
- The authors argue LLMs function mainly as fast, associative pattern predictors, akin to System 1 thinking.
- Equipping AI with explicit reasoning capabilities, the piece suggests, could be necessary for trustworthy scientific or medical breakthroughs.
Source: technologyreview.com — Thore Graepel, 2026-10-02
Published there as: “Don’t be fooled—LLMs don’t reason”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.